Abstract: Forecasting in the environment of physical culture and sports is the most important element of planning the training process and sports achievements. The active development of computer technology and applied artificial intelligence allows the use of modern information technology and powerful mathematical apparatus in the field of forecasting. The article provides an example of a software system for predicting sports results in individual and team sports. The structure of the system and the basic principle of its operation are described. One of the modules of the system, which implements the neural network of vector quantization and solves the problem of data clustering, is considered in more detail. Clustering method is one of the most effective ways to predict sports events. The detailed algorithm of the module functioning is given. The results of experiments using this module are analyzed. The Boxing match between British boxer Anthony Joshua and Russian boxer Alexander Povetkin for the title of world champion was chosen as the studied object. The main predicted event was the determination of the direct outcome of the match. The description of one of the variants of training sample preparation is given. The evaluation of the effectiveness of the formed training sample and the results of experiments are presented. The use of a module that implements a neural network of vector quantization to predict the results of a Boxing match can be considered successful. The prototype of the proposed software system based on some models of artificial neural networks was developed in MATLAB. The system, which includes the module presented in the article, can be used not only for forecasting individual and team competitions, like bookmaking applications, but also for planning individual achievements, for example, in swimming, athletics, etc., taking into account anthropometric data and control results shown by athletes during control trainings and special tests.
Index terms: sports forecasting, software system, neural network, training sampling, vector quantization, LVQ network.